Video model

MARC

MARC is a 4B parameter video understanding model that uses memory augmented reinforcement learning to compress video tokens, cutting computation while preserving accuracy for long video analysis and captioning tasks.

Type
Video
Weights
Open

More from Memories.ai

4

About Memories.ai

1

Memories.ai is an AI research company that builds a "Large Visual Memory Model" (video AI) enabling machines to index, search, and act on large-scale video/camera footage.